Update app.py

app.py

@@ -67,6 +67,107 @@ if torch.cuda.is_available():
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.use_default_system_prompt = False
 
+## trust-game-llama-2-7b-chat
+# app.py 
+@spaces.GPU
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    # system_prompt: str,
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50,
+    repetition_penalty: float = 1.2,
+) -> Iterator[str]: # Change return type hint to Iterator[str]
+    
+
+    # Construct the input prompt using the functions from the system_prompt_config module
+    input_prompt = construct_input_prompt(chat_history, message, personalized_data)
+
+    # Use the global variable to store the chat history
+    # global global_chat_history
+
+    conversation = []
+
+    # Move the condition here after the assignment
+    if input_prompt:
+        conversation.append({"role": "system", "content": input_prompt})
+
+    # Convert input prompt to tensor
+    input_ids = tokenizer(input_prompt, return_tensors="pt").to(model.device)
+
+
+    for user, assistant in chat_history:
+        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+    conversation.append({"role": "user", "content": message})
+
+    input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
+    if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
+        input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
+        gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
+    input_ids = input_ids.to(model.device)
+
+    # Set up the TextIteratorStreamer
+    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
+    
+    # Set up the generation arguments
+    generate_kwargs = dict(
+        {"input_ids": input_ids},
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
+        do_sample=True,
+        top_p=top_p,
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+        repetition_penalty=repetition_penalty,
+    )
+
+    # Start the model generation thread
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    # Yield generated text chunks
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
+
+    # Update the global_chat_history with the current conversation
+    # global_chat_history.append({
+    #    "message": message,
+    #    "chat_history": chat_history,
+    #    "system_prompt": input_prompt,
+    #    "output": outputs[-1],  # Assuming you want to save the latest model output
+    # })
+    
+    # The modification above starting with "global_chat.history.append" introduces a global_chat_history variable to store the chat history globally. 
+    # The save_chat_history function is registered to be called when the program exits 
+    # using atexit.register(save_chat_history). 
+    # It saves the chat history to a JSON file named "chat_history.json". 
+    # The generate function is updated to append the current conversation to global_chat_history 
+    # after generating each response.
+
+        
+    #gr.Markdown(DESCRIPTION)
+    #gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
+    ## gradio-chatbot-read-query-param
+
+chat_interface = gr.ChatInterface(
+fn=generate,
+theme="soft",
+retry_btn=None,
+clear_btn=None,
+undo_btn=None,
+chatbot=gr.Chatbot(avatar_images=('user.png', 'bot.png'), bubble_full_width = False), 
+examples=[
+    ["Can you explain the rules very briefly again?"],
+    ["How much should I invest in order to win?"],
+    ["What happened in the last round?"],
+    ["What is my probability to win if I do not share anything?"],
+],
+)
 
 with gr.Blocks(css="style.css") as demo:
     ## gradio-chatbot-read-query-param   
@@ -78,6 +179,11 @@ with gr.Blocks(css="style.css") as demo:
                 return url_params;
                 }
             """
+    def get_session_index(chat_history, url_params):
+        if chat_history and bool(chat_history[-1][0].strip()):
+            session_index = url_params.get('session_index')
+            print(session_index)
+        return session_index
     
     def fetch_personalized_data(session_index):
         # Connect to the database
@@ -125,116 +231,6 @@ with gr.Blocks(css="style.css") as demo:
             print(f"Error: {err}")
             return None
     
-    ## trust-game-llama-2-7b-chat
-    # app.py
-    def construct_input_prompt(chat_history, message, personalized_data):
-        input_prompt = f"<s>[INST] <<SYS>>\n{get_default_system_prompt(personalized_data)}\n<</SYS>>\n\n "
-    
-        for user, assistant in chat_history:
-            input_prompt += f"{user} [/INST] {assistant} <s>[INST] "
-    
-        input_prompt += f"{message} [/INST] "
-    
-        return input_prompt
-
-    
-    ## trust-game-llama-2-7b-chat
-    # app.py 
-    @spaces.GPU
-    def generate(
-        message: str,
-        chat_history: list[tuple[str, str]],
-        # system_prompt: str,
-        max_new_tokens: int = 1024,
-        temperature: float = 0.6,
-        top_p: float = 0.9,
-        top_k: int = 50,
-        repetition_penalty: float = 1.2,
-    ) -> Iterator[str]: # Change return type hint to Iterator[str]
-        
-        # Fetch personalized data
-        url_params = get_window_url_params()
-        session_index = get_session_index(chat_history, url_params)
-        personalized_data = fetch_personalized_data(session_index)
-    
-         # Construct the input prompt using the functions from the system_prompt_config module
-        input_prompt = construct_input_prompt(chat_history, message, personalized_data)
-    
-        # Use the global variable to store the chat history
-        # global global_chat_history
-    
-        conversation = []
-    
-        # Move the condition here after the assignment
-        if input_prompt:
-            conversation.append({"role": "system", "content": input_prompt})
-    
-        # Convert input prompt to tensor
-        input_ids = tokenizer(input_prompt, return_tensors="pt").to(model.device)
-    
-    
-        for user, assistant in chat_history:
-            conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
-        conversation.append({"role": "user", "content": message})
-    
-        input_ids = tokenizer.apply_chat_template(conversation, return_tensors="pt")
-        if input_ids.shape[1] > MAX_INPUT_TOKEN_LENGTH:
-            input_ids = input_ids[:, -MAX_INPUT_TOKEN_LENGTH:]
-            gr.Warning(f"Trimmed input from conversation as it was longer than {MAX_INPUT_TOKEN_LENGTH} tokens.")
-        input_ids = input_ids.to(model.device)
-    
-        # Set up the TextIteratorStreamer
-        streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
-        
-        # Set up the generation arguments
-        generate_kwargs = dict(
-            {"input_ids": input_ids},
-            streamer=streamer,
-            max_new_tokens=max_new_tokens,
-            do_sample=True,
-            top_p=top_p,
-            top_k=top_k,
-            temperature=temperature,
-            num_beams=1,
-            repetition_penalty=repetition_penalty,
-        )
-    
-        # Start the model generation thread
-        t = Thread(target=model.generate, kwargs=generate_kwargs)
-        t.start()
-    
-        # Yield generated text chunks
-        outputs = []
-        for text in streamer:
-            outputs.append(text)
-            yield "".join(outputs)
-    
-        # Update the global_chat_history with the current conversation
-        # global_chat_history.append({
-        #    "message": message,
-        #    "chat_history": chat_history,
-        #    "system_prompt": input_prompt,
-        #    "output": outputs[-1],  # Assuming you want to save the latest model output
-        # })
-     
-        # The modification above starting with "global_chat.history.append" introduces a global_chat_history variable to store the chat history globally. 
-        # The save_chat_history function is registered to be called when the program exits 
-        # using atexit.register(save_chat_history). 
-        # It saves the chat history to a JSON file named "chat_history.json". 
-        # The generate function is updated to append the current conversation to global_chat_history 
-        # after generating each response.
-    
-            
-        #gr.Markdown(DESCRIPTION)
-        #gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
-        ## gradio-chatbot-read-query-param
-
-    def get_session_index(chat_history, url_params):
-        if chat_history and bool(chat_history[-1][0].strip()):
-            session_index = url_params.get('session_index')
-            print(session_index)
-        return session_index
-
     ## trust-game-llama-2-7b-chat
     # app.py 
     def get_default_system_prompt(personalized_data):
@@ -263,21 +259,20 @@ with gr.Blocks(css="style.css") as demo:
         print(DEFAULT_SYSTEM_PROMPT)
         return DEFAULT_SYSTEM_PROMPT
     
-    chat_interface = gr.ChatInterface(
-    fn=generate,
-    theme="soft",
-    retry_btn=None,
-    clear_btn=None,
-    undo_btn=None,
-    chatbot=gr.Chatbot(avatar_images=('user.png', 'bot.png'), bubble_full_width = False), 
-    examples=[
-        ["Can you explain the rules very briefly again?"],
-        ["How much should I invest in order to win?"],
-        ["What happened in the last round?"],
-        ["What is my probability to win if I do not share anything?"],
-    ],
-    )
+    ## trust-game-llama-2-7b-chat
+    # app.py
+    def construct_input_prompt(chat_history, message, personalized_data):
+        input_prompt = f"<s>[INST] <<SYS>>\n{get_default_system_prompt(personalized_data)}\n<</SYS>>\n\n "
+        for user, assistant in chat_history:
+            input_prompt += f"{user} [/INST] {assistant} <s>[INST] "
+        input_prompt += f"{message} [/INST] "
+        return input_prompt
     
+    # Fetch personalized data
+    url_params = get_window_url_params()
+    session_index = get_session_index(chat_history, url_params)
+    personalized_data = fetch_personalized_data(session_index)
+
     chat_interface.render()
     #gr.Markdown(LICENSE)
 
@@ -290,4 +285,3 @@ if __name__ == "__main__":
 # Register the function to be called when the program exits
 # atexit.register(save_chat_history)
     
-